Memory System Adaptive Entropy Coding Latency Reduction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Adaptive entropy coding in data compression systems experiences high latency and low throughput due to extensive processing required for updating frequency tables, leading to increased circuit size and critical path lengths.
Innovation Solution
The system employs a frequency value reduction conversion process that starts without waiting for the update of the frequency table, and frequency entries are divided into groups for simplified sorting and swapping between adjacent groups, reducing the number of comparators needed and shortening the critical path.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Loss of time
If adaptive entropy coding is used to compress data with shorter latency, then the coding speed improves, but the circuit size and critical path length increase due to extensive frequency table update processing
Solution Approach 1:
The patent divides the frequency table update process into two independent segments: (1) updating frequency values by adding occurrence counts, and (2) performing reduction conversion on the frequency values. These segmented operations can be executed in parallel, reducing the critical path length and circuit complexity while maintaining adaptive coding performance.
Solution Approach 2:
The patent performs frequency value reduction conversion in advance, before the frequency table update is complete. By initiating the reduction conversion process preliminarily and independently, the system avoids waiting for the complete update cycle, thereby reducing latency and critical path without requiring additional circuit complexity.
2Manufacturing precision
If frequency table update processing is performed extensively to maintain compression efficiency, then the compression ratio improves, but the throughput decreases due to processing delays
Solution Approach 1:
The patent ensures continuous useful action by allowing the reduction conversion process to proceed continuously and independently from the frequency table update process. This continuity maintains compression efficiency while preventing throughput degradation, as the conversion operation never waits for update completion.
Solution Approach 2:
By performing reduction conversion preliminarily and independently, the system prepares compressed data for subsequent processing without being blocked by ongoing frequency table updates. This preliminary action maintains both compression efficiency and high throughput.
3Measurement precision
If the frequency value reduction conversion waits for the frequency table update to complete, then the accuracy of compression is maintained, but the processing time increases
Solution Approach 1:
The patent performs reduction conversion preliminarily based on available frequency data without waiting for update completion. This preliminary conversion maintains sufficient accuracy for compression while significantly reducing processing time, as the operation proceeds with partial data that is continuously refined.
Solution Approach 2:
By segmenting the conversion process from the update process, the patent allows accuracy to be maintained through iterative refinement while avoiding time loss from sequential waiting. The segmented architecture enables parallel execution without compromising final compression accuracy.
Data Source
AI summary
According to one embodiment, a memory system includes a frequency value update unit, a first intra-group rearranging unit, a second intra-group rearranging unit, and a routing unit. The frequency value update unit adds a first value to each of one or more first frequency values that are associated with one or more first symbols, respectively, in one or more entries in a table. The first intra-group rearranging unit rearranges first entries, which belong to a first group, using frequency values in the first entries. The second intra-group rearranging unit rearranges second entries, which belong to a second group lower than the first group, using frequency values. The routing unit rearranges at least one of the first entries and at least one of the second entries.